VLUC: An Empirical Benchmark for Video-Like Urban Computing on Citywide Crowd and Traffic Prediction

Jiang, Renhe; Cai, Zekun; Wang, Zhaonan; Yang, Chuang; Fan, Zipei; Song, Xuan; Tsubouchi, Kota; Shibasaki, Ryosuke

doi:10.48550/arxiv.1911.06982

Cited by 4 publications

(3 citation statements)

References 48 publications

(49 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in Figure 4, this spatio-temporal attention module is further divided into two sub-modules, a spatial attention 3D convolution. More works [10,41,42] have shown that a stack of 2D convolutions can effectively model nearby spatial dependencies, as well as distant spatial dependencies, in traffic data. However, 2D convolutions lack the ability to capture the temporal dependencies in multiple traffic congestion matrices.…”

Section: Sta Modulementioning

confidence: 99%

AST3DRNet: Attention-Based Spatio-Temporal 3D Residual Neural Networks for Traffic Congestion Prediction

Li,

Dai,

Huang

et al. 2024

Sensors

View full text Add to dashboard Cite

Traffic congestion prediction has become an indispensable component of an intelligent transport system. However, one limitation of the existing methods is that they treat the effects of spatio-temporal correlations on traffic prediction as invariable during modeling spatio-temporal features, which results in inadequate modeling. In this paper, we propose an attention-based spatio-temporal 3D residual neural network, named AST3DRNet, to directly forecast the congestion levels of road networks in a city. AST3DRNet combines a 3D residual network and a self-attention mechanism together to efficiently model the spatial and temporal information of traffic congestion data. Specifically, by stacking 3D residual units and 3D convolution, we proposed a 3D convolution module that can simultaneously capture various spatio-temporal correlations. Furthermore, a novel spatio-temporal attention module is proposed to explicitly model the different contributions of spatio-temporal correlations in both spatial and temporal dimensions through the self-attention mechanism. Extensive experiments are conducted on a real-world traffic congestion dataset in Kunming, and the results demonstrate that AST3DRNet outperforms the baselines in short-term (5/10/15 min) traffic congestion predictions with an average accuracy improvement of 59.05%, 64.69%, and 48.22%, respectively.

show abstract

Section: Sta Modulementioning

confidence: 99%

AST3DRNet: Attention-Based Spatio-Temporal 3D Residual Neural Networks for Traffic Congestion Prediction

Li,

Dai,

Huang

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

“…Many studies have been conducted to apply CNNs to capture the spatial features of traffic networks [10], [11]. CNN-based models are suitable for solving the problems that involve modeling the Euclidean correlations among different regions [12], [13]. These studies converted traffic networks to regular grids because the CNNs can only process Euclideanstructured data.…”

Section: A Gdl In Transportationmentioning

confidence: 99%

Optimizing Systemic Redundancy of Traffic Sensor Networks While Maintaining Resilience: New Evidence From Using Graph Learning

Tang

Wei

et al. 2023

IEEE Systems Journal

View full text Add to dashboard Cite

The optimization of systemic redundancy by minimizing the sensor quantity can improve the efficiency of sensor networks and save costs. However, from the perspective of risk management, this redundancy reduction can also bring a significant loss in the overall network resilience because the less the systemic redundancy is, the fewer backup components in the network when shocks hit and, therefore, the less overall resilience. In this paper, we investigate this intractable dilemma and attempt to pinpoint the tradeoff point for a city-scale automatic number plate recognition (ANPR) system in Cambridge, UK. By developing a two-stage graph deep learning (GDL) model, we first optimize the layout of the ANPR system to reduce redundancy and find its efficiency profile. Next, we study what effects this redundancy reduction can bring to the overall resilience, as the overall observability drops with the reduction in the number of sensors and find an optimal balance. The results show that our approach can effectively optimize the system's redundancy by using only 47% of the original sensors to reconstruct the full picture with a mean absolute error (MAE) of only 11.18 and a root mean square error (RMSE) of 19.49; most importantly, the overall system resilience is maintained at 70% in the meantime. This paper provides an alternative perspective for dealing with the well-known 'efficiency-resilience' dilemma and offers new evidence to enable better decision and policy making for city managers and planners in local authorities.

show abstract

“…Many studies have proposed different methods for the task of crowd prediction, such as DCRNN [ 28 ], SRCNs [ 29 ], and multitask-net [ 30 ]. VLUC [ 31 ], PCRN [ 32 ], and PDB-ConvLSTM [ 33 ] use CNNs to process recent, near, and far data, respectively, and treat each timestamp as the equivalent convolution channel. STGCN [ 17 ], MRGCN [ 34 ], and ST-MGCN [ 34 ] fit a graph to the road structure and use convolution to learn temporal correlations.…”

Section: Related Workmentioning

confidence: 99%

Discovering Key Sub-Trajectories to Explain Traffic Prediction

Wang

Fan

Chen

et al. 2022

Sensors

View full text Add to dashboard Cite

Flow prediction has attracted extensive research attention; however, achieving reliable efficiency and interpretability from a unified model remains a challenging problem. In the literature, the Shapley method offers interpretable and explanatory insights for a unified framework for interpreting predictions. Nevertheless, using the Shapley value directly in traffic prediction results in certain issues. On the one hand, the correlation of positive and negative regions of fine-grained interpretation areas is difficult to understand. On the other hand, the Shapley method is an NP-hard problem with numerous possibilities for grid-based interpretation. Therefore, in this paper, we propose Trajectory Shapley, an approximate Shapley approach that functions by decomposing a flow tensor input with a multitude of trajectories and outputting the trajectories’ Shapley values in a specific region. However, the appearance of the trajectory is often random, leading to instability in interpreting results. Therefore, we propose a feature-based submodular algorithm to summarize the representative Shapley patterns. The summarization method can quickly generate the summary of Shapley distributions on overall trajectories so that users can understand the mechanisms of the deep model. Experimental results show that our algorithm can find multiple traffic trends from the different arterial roads and their Shapley distributions. Our approach was tested on real-world taxi trajectory datasets and exceeded explainable baseline models.

show abstract

VLUC: An Empirical Benchmark for Video-Like Urban Computing on Citywide Crowd and Traffic Prediction

Cited by 4 publications

References 48 publications

AST3DRNet: Attention-Based Spatio-Temporal 3D Residual Neural Networks for Traffic Congestion Prediction

AST3DRNet: Attention-Based Spatio-Temporal 3D Residual Neural Networks for Traffic Congestion Prediction

Optimizing Systemic Redundancy of Traffic Sensor Networks While Maintaining Resilience: New Evidence From Using Graph Learning

Discovering Key Sub-Trajectories to Explain Traffic Prediction

Contact Info

Product

Resources

About